论文部分内容阅读
软测量技术可以有效解决复杂工业过程中一些重要参量难以由硬件在线检测的问题,由于化工过程具有连续性和累积性等特点,若采用传统的软测量建模方法往往会忽略信号的时间累积作用从而导致预测误差较大。针对上述问题,提出了基于改进的过程神经网络(PNN)的软测量建模方法。首先采用移动窗技术来确定包含过程正常运行大部分信息的时间序列,然后利用改进的PNN建立软测量模型并对主导变量进行连续预测,最后对软仪表进行校正以实现连续高精度预测。以某工厂高密度聚乙烯装置为例,验证了该方法具有较高的预测精度和跟踪性能,这对于工业过程的控制优化操作具有重要的应用价值。
Soft measurement technology can effectively solve the problem that some important parameters in complex industrial processes are difficult to be detected online by hardware. Due to the characteristics of continuity and accumulation of chemical processes, traditional time-accumulated signals tend to be ignored if traditional soft-sensing modeling methods are used Resulting in a larger prediction error. In view of the above problems, a soft-sensing modeling method based on improved process neural network (PNN) is proposed. Firstly, the moving window technique is used to determine the time series that contains most of the information of the normal operation of the process. Then, the improved PNN is used to establish the soft-sensing model and predict the dominant variables continuously. Finally, the soft instruments are calibrated to achieve continuous high-precision prediction. Taking a high-density polyethylene plant in a factory as an example, the method has been proved to have high prediction accuracy and tracking performance, which has important application value for controlling and optimizing industrial processes.